188-OR: Assessing the Validity of Diagnosis Codes for Gestational Diabetes Ascertainment

Background: Researchers often rely on diagnosis codes from billing/administrative data to ascertain diagnoses, including gestational diabetes (GDM). We assessed the validity of relying on these diagnosis codes for GDM ascertainment, using laboratory criteria for GDM as the reference. Methods: We studied women seen at our hospital for prenatal care between 1998-2016. Women underwent GDM screening using a 50-gram glucose loading test (GLT) followed by a diagnostic 100-gram 3-hour oral glucose tolerance test (GTT) if the GLT result was ≥140 mg/dl. We defined GDM using National Diabetes Data Group (NDDG) criteria, which was in clinical use during the study period. We calculated the sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) for ICD code-based ascertainment of GDM using codes for GDM (ICD-9: 648.8; ICD-10: O24.4) recorded in administrative data from inpatient and outpatient encounters. In a sensitivity analysis we included additional codes that broadly represent diabetes and abnormal glucose diagnoses (ICD-9: 249, 250, 648.0, 790.2; ICD-10: O24.0-3, O24.8-9, E08-E13, R73.0, R73.9). In a secondary analysis, we defined GDM using Carpenter-Coustan (CC) criteria. Results: Out of 49195 pregnancies, 1258 (2.6%) met NDDG criteria and 2039 (4.1%) met CC criteria based on GLT and GTT values. In the same cohort, 589 (1.2%) had GDM ICD-9/10 codes. Using NDDG criteria as the reference, we observed sensitivity of 35.2% and specificity of 99.7% for these codes. This resulted in a PPV of 75.2% and an NPV of 98.3%. In our sensitivity analysis, 700 (1.4%) pregnancies had at least one of the codes on the expanded list. The sensitivity (37.4%) and specificity (99.5%) of this set of codes was similar to the more restricted set. Results were similar for GDM defined by CC criteria. Conclusion: Using diagnosis codes from administrative data to ascertain GDM resulted in underestimation of cases. One in 4 cases identified by these codes could not be confirmed with laboratory data. Disclosure S. Hsu: None. K. Mathieu: None. A. Kaimal: None. R. Thadhani: None. K. James: None. C.E. Powe: None. Funding Massachusetts General Hospital